Molecular Phylogenetics of the Bee-Eaters (Aves: Meropidae) Based on Nuclear and Mitochondrial DNA Sequence Data

Molecular Phylogenetics and Evolution 45 (2007) 23–32 www.elsevier.com/locate/ympev

Molecular phylogenetics of the bee-eaters (Aves: Meropidae) based on nuclear and mitochondrial DNA sequence data

Ben D. Marks a,*, Jason D. Weckstein b, Robert G. Moyle c

a Department of Biological Sciences and Museum of Natural Science, Louisiana State University, 119 Foster Hall, Baton Rouge, LA 70808, USA b Zoology Department, Field Museum of Natural History, 1400 South Lake Shore Drive, Chicago, IL 60605, USA c Department of Ornithology, American Museum of Natural History; Natural History Museum and Biodiversity Research Center, University of Kansas, KS, USA

Received 19 October 2006; revised 27 June 2007; accepted 3 July 2007 Available online 18 July 2007

Abstract

The bee-eaters (family Meropidae) comprise a group of brightly colored, but morphologically homogeneous, birds with a wide variety of life history characteristics. A phylogeny of bee-eaters was reconstructed using nuclear and mitochondrial DNA sequence data from 23 of the 25 named bee-eater species. Analysis of the combined data set provided a well-supported phylogenetic hypothesis for the family. Nyctiornis is the sister taxon to all other bee-eaters. Within the genus Merops, we recovered two well-supported clades that can be broadly separated into two groups along geographic and ecological lines, one clade with mostly African resident species and the other clade containing a mixture of African and Asian taxa that are mostly migratory species. The clade containing resident African species can be further split into two groups along ecological lines by habitat preference into lowland forest specialists and montane forest and forest edge species. Intraspeciﬁc sampling in several of the taxa revealed moderate to high (3.7–6.5%, ND2) levels of divergence in the resident taxa, whereas the lone migratory taxon showed negligible levels of intraspeciﬁc divergence. This robust molecular phylogeny provides the phylogenetic framework for future comparative tests of hypotheses about the evolution of plumage patterns, sociality, migration, and delayed breeding strategies. Ó 2007 Elsevier Inc. All rights reserved.

Keywords: Bee-eater; Meropidae; Systematics; Merops; Nyctiornis; Meropogon

1. Introduction and life history traits. Bee-eaters are distributed throughout the Old World, found in habitats ranging from the dri- The bee-eaters (family Meropidae) comprise a group of est of deserts to the wettest of rainforests. The family spans 25 species (Dickinson, 2003) of brightly colored, but mor- a spectrum ranging from sedentary tropical rainforest spe- phologically homogeneous, birds. As their common name cies to species that migrate between the temperate zone and suggests, bee-eaters feed primarily on bees and other the tropics for breeding and wintering. Some species of bee- Hymenoptera, which are usually plucked from the air in eaters nest solitarily, whereas others nest in large breeding gracefully aggressive foraging forays. Bee-eaters generally colonies. The young of some species delay breeding in their possess a conserved body plan with a long decurved bill, ﬁrst year to serve as ‘‘helpers’’ at the nests of their parents long slender wings, a long tail (sometimes with elongated (Fry, 1972; Hegner et al., 1982). The remarkable level of central rectrices), and usually bold green or red base color- heterogeneity of behaviors (migratory, social, and breed- ation. The low level of morphological diversity in the fam- ing) among the bee-eaters is ideal for comparative study, ily is in stark contrast to their high diversity in ecological which requires a well-resolved phylogeny. However, until now there has been only a single formal analysis of bee- * Corresponding author. Fax: +1 225 578 3075. eater phylogenetic relationships (Burt, 2004) and no quan- E-mail address: [email protected] (B.D. Marks). titative assessment of support for relationships.

Most authors recognize three genera in the family: Nyct- suggests that the phylogenetic signal in coloration charac- iornis, which includes two species of humid forest special- ters may be weak. ists inhabiting the rainforests of India and SE Asia, To test these proposed relationships among bee-eaters, Meropogon, a monotypic genus inhabiting forest edge we reconstructed a molecular phylogeny for the family and clearings in Sulawesi, and Merops, which includes 22 based on mitochondrial and nuclear DNA sequence data. species distributed across a variety of habitats throughout In the process, we resolve many phylogenetic and taxo- the Old World. Two of the most thorough investigations nomic uncertainties in the family and provide a phyloge- of the relationships among bee-eater species are those of netic framework for future comparative tests of von Boetticher (1951) and Fry (1969). Both studies were hypotheses about the evolution of plumage patterns, soci- based primarily on plumage (feather structure and color ality, migration and delayed breeding strategies. pattern) and morphometric features, although Fry (1969, 1984) also considered ecology and biogeography in produc- 2. Materials and methods ing his hypothesis of relationships. Although generic names applied in these revisions differed, the classifications were 2.1. Taxon and character sampling largely congruent, differing primarily in the placement of the African forest species (Merops muelleri, M. gularis, Ingroup sampling for our study (Table 1) included rep- and M. breweri). von Boetticher (1951) considered M. resentatives of 23 out of the 25 named species (Dickinson, breweri, a large and mostly solitary species, to be closely 2003), missing only Nycitornis athertoni and Merops revolii. related to the SE Asian N. amicta, N. athertoni,andMero- We also sampled two subspecies each for four of the species pogon forsteni and placed all four taxa in the genus Nyctior- (Table 1). Tissue samples were available for most taxa, but nis. The remaining African forest specialists (M. gularis for some taxa, only museum study-skin toepads or blood and M. muelleri) were placed in a separate genus Meropis- samples were available (Table 1). Outgoup taxa were cus. Fry (1969), using primarily biogeographic informa- drawn from other Coraciiform families; kingfishers, mot- tion, treated M. breweri as a member of Merops, but mots and todies. Genomic DNA was extracted from mus- later conceded that M. breweri may indeed be more closely cle tissue using the Qiagen DNeasy tissue kit following the related to Meropogon (Fry, 1984). Fry (1969, 1984) manufacturer’s protocol (Qiagen, Valencia, California). remained hesitant to group all of the lowland forest taxa Protocols for DNA extraction from museum skins gener- into Nyctiornis. These early hypotheses of relationships ally followed those for fresh tissues, with the addition of among bee-eater species, based on phenetic analyses of 10 ll of 1 M dithiothreitol solution. The protocols also morphological characters and life history data, were required special handling of samples and prevention of recently revisited by Burt (2004) in a comprehensive cladis- contamination, as described by Mundy et al. (1997). We tic analysis of 30 plumage-based and morphometric char- extracted, amplified, and sequenced all bee-eater tissue acters. The relationships inferred from Burt’s (2004) samples in the molecular labs at the LSU Museum of Nat- analysis were generally consistent with previous studies, ural Science and the Pritzker Laboratory at The Field but added M. hirundineus and M. boehmi to the list of taxa Museum, whereas the toepad samples were extracted, with enigmatic relationships. Burt (2004) found support for amplified, and sequenced in the Cullman Molecular Labo- a sister relationship between the African M. hirundineus ratory at the American Museum of Natural History. and the Australian M. ornatus, a relationship not sup- From the tissue samples we collected DNA sequence ported by either Fry (1969, 1984) or von Boetticher data from the entire second subunit of the mitochondrial (1951); both of whom place M. hirundineus in a superspe- nicotinamide adenine dinucleotide dehydrogenase gene cies consisting of other African bee-eaters (M. oreobates/ (ND2, 1041 bp) and two nuclear introns: the fifth intron pusillus/variegatus). In Burt’s (2004) four equally parsimo- of the nuclear b-fibrinogen gene (BFib5, 604 bp) and the nious trees, M. boehmi exhibited sister relationships with fifth intron of the transforming growth factor, b2 gene M. superciliosus/philippinus or M. viridis/leschenaulti, (TGFb2, 566 bp). For toepad samples we sequenced the whereas the earlier treatments placed M. boehmi close to entire second subunit of the ND2 gene (primers available the widespread M. orientalis (Fry, 1969) or the African upon request). We used primers L5215 (Hackett, 1996), M. albicollis (von Boetticher, 1951). The plumage charac- H6313, L5758, and H5766 (Johnson and Sorenson, 1998) ters used by Burt (2004) did not lend themselves to quanti- to amplify and sequence the ND2 gene, primers FIB5L tative analysis such as bootstrapping, and as a result there and FIB5H (Driskell and Christidis, 2004) to amplify and is no measure of support for the nodes depicted in his phy- sequence the Fib5 intron, and primers TGF5 and TGF6 logenetic hypothesis. Furthermore, plumage characters, (Primmer et al., 2002) to amplify and sequence the TGFb2 such as those used by Burt (2004), are known to evolve rap- intron. We purified PCR products with Perfectprep PCR idly in other birds (Omland and Hofmann, 2006) and a cleanup kits (Eppendorf) or with Exonuclease and Shrimp number of studies (Johnson, 1999; Crochet et al., 1999; Alkaline Phosphotase enzymatic reactions (United States Johnson and Lanyon, 2000; Omland and Lanyon, 2000; Biochemical). Sequencing of purified PCR products was Armenta et al., 2005; Weckstein, 2005) have found high performed with BigDye Terminator Cycle Sequencing Kits levels of homoplasy in plumage coloration patterns, which (Applied Biosystems). Primers used for PCR were also used B.D. Marks et al. / Molecular Phylogenetics and Evolution 45 (2007) 23–32 25

Table 1 Voucher information for species included in the study Genus Species Subspecies Distribution Habitat R/Ma Voucher information Nyctiornis amicta Asia Forest R LSUMNS B47037 Meropogon forsteni Asia Forest R AMNH 299316b Meropogon forsteni Asia Forest R AMNH 299315b Merops leschenaulti Asia Forest/savanna R AMNH 409286b Merops leschenaulti Asia Forest/savanna R AMNH 812115b Merops boehmi Africa Savanna R ZMUC 114552 Merops gularis gularis Africa Forest R LSUMNS B39368 Merops gularis australis Africa Forest R ANSP 11535 Merops pusillus pusillus Africa Savanna R LSUMNS B39280 Merops pusillus meridionalis Africa Savanna R MBM 11049 Merops viridis viridis Asia Forest/savanna R LSUMNS B51022 Merops viridis americanus Asia Forest/savanna R FMNH 358335 Merops persicus persicus Africa/Asia/Europe Savanna M LSUMNS B46361 Merops apiaster Africa/Asia/Europe Savanna M LSUMNS B35153 Merops apiaster Africa/Asia/Europe Savanna M MBM 11515 Merops muelleri mentalis Africa Forest R LSUMNS B45318 Merops muelleri muelleri Africa Forest R ANSP 11434 Merops bullocki bullocki Africa Savanna R LSUMNS B39305 Merops albicollis Africa Forest M LSUMNS B39423 Merops ornatus Australia Forest M ANSP 11226 Merops nubicus nubicoides Africa Savanna M UWBM 57051 Merops superciliosus superciliosus Europe/Africa Savanna M FMNH 384686 Merops oreobates Africa Forest R FMNH 384825 Merops variegatus loringi Africa Forest/savanna R FMNH 346214 Merops bullockoides Africa Savanna M MBM 11950 Merops hirundineus Africa Savanna M MBM 8013 Merops phillipinus Asia Savanna M C-40531c Merops malimbicus Africa Savanna M NMNH B16241 Merops orientalis Africa/Asia Savanna R NMNH B5669 Merops breweri Africa Forest R NMNH B16221 Actenoides concretus LSUMNS B36383 Momotus momota AMNH 12359 Todus angustirostris AMNH 6949 Each species is broadly categorized by distribution, habitat preference (forest or savanna), and as either resident (R) or migratory (M). Taxonomy follows Dickinson (2003). Note: AMNH, American Museum of Natural History; ANSP, Academy of Natural Sciences Philadelphia; FMNH, Field Musuem of Natural History; LSUMNS, Louisiana State University Museum of Natural Science; MBM, Marjorie Barrick Museum of Natural History UNLV; NMNH, Smithsonian Institution National Museum of Natural History; UWBM, University of Washington Burke Museum; ZMUC, Zoological Museum University of Copenhagen. a R/M—R, resident; M, migratory. b DNA sample collected from study-skin toe pad. c No voucher, blood sample.

for cycle sequencing reactions, resulting in bi-directional three diﬀerent ILD tests including one between ND2 and sequence for all taxa. Cycle sequencing products were run the two nuclear loci combined, one between the three sepa- on an ABI Prism 3100 or 3730 automated DNA sequencer rate partitions (ND2, BFib5, and TGFb2), and one between (Perkin-Elmer Applied Biosystems). The computer pro- the two nuclear partitions (BFib5 and TGFb2). For each of gram Sequencer 4.1 (Genecodes) was used to reconcile these ILD tests, we removed taxa for which we only have chromatograms of complementary fragments and align ND2 data and we excluded all characters that were not par- sequences across taxa. All of the sequence data generated simony informative from the analyses (Cunningham, 1997). in this study have been submitted to GenBank (Accession PAUP*4.0b10 was used to test the base composition of each Nos. EU21508–EU21595). gene using a v2 analysis of base frequencies across taxa. To visualize the degree of divergence between ingroup and out- 2.2. Phylogenetic analysis group taxa, the extent of saturation, and relative substitution rates between the nuclear and mitochondrial data, we We used the partition homogeneity test (ILD statistic, plotted the uncorrected distance (p-distance) and maximum Farris et al., 1994, 1995) as implemented in PAUP* (version likelihood transformed distances of ND2 and the faster of 4.0b10; Swoﬀord, 2002) to compare phylogenetic signal and the two nuclear genes, TGFb2 against those of FIB5 for to test for incongruence between data partitions. We ran all pairwise comparisons of taxa. 26 B.D. Marks et al. / Molecular Phylogenetics and Evolution 45 (2007) 23–32

Maximum likelihood (ML) and maximum parsimony (MP) analyses were performed for the combined data set using PAUP* 4.0b10 (Swofford, 2002). Heuristic searches employed TBR branch-swapping and 100 random taxon addition sequences (10 addition sequences for ML searches). For each likelihood analysis, we used the Akaike Information Criterion in Modeltest 3.5 (Posada and Crandall, 1998) to determine parameter estimates and the best fit model of molecular evolution. Support for nodes in the ML and MP trees were assessed by analysis of 100 non-parametric bootstrap replicates (Felsenstein, 1985). For the combined data set, we implemented a mixed model approach to account for the potential difference in evolutionary model parameters between data partitions (Nylander et al., 2004). We used MrModeltest2.2 (Nyland- er, 2004) to determine the evolutionary model appropriate for each partition and chose among partitioning schemes using Bayes factors (see Brandley et al., 2005 for a thorough discussion of the methodology and rationale). Bayes factors were calculated using the harmonic mean from the sump command within MrBayes. A difference of 2ln Bayes factor >10 was used as the minimum value to discriminate between partitioning schemes. MrBayes 3.1.1 (Ronquist and Huelsenbeck, 2003) was used to estimate model parameters from the data and to evaluate support for specific relationships in the phylogeny. All parameters (except topology and branch lengths) were unlinked between partitions. We ran four Markov chains for 10 million generations as well as two shorter 2 million generation runs. The shorter runs were used to ensure that our analyses were not stuck in local optima (Huelsenbeck and Bollback, 2001) and to evaluate stationarity, the condition in which Fig. 1. Comparison of pairwise divergences (uncorrected p-distance and maximum likelihood distance) among nuclear and mitochondrial gene trees are being sampled according to their posterior proba- regions sequenced for this study. BFib5 distances are plotted on the x-axis, bilities. Stationarity was judged by visually inspecting plots the mitochondrial ND2 and the nuclear TGFb2 are plotted on the y-axis. of likelihood scores. Markov chains were sampled every (A) Uncorrected proportional distances. (B) Maximum likelihood distances 1000 generations, yielding 10,000 parameter point esti- determined using the models chosen by Modeltest. ND2: Model = G- mates. These subsamples, minus the burn-in generations, TR + I + G, base frequencies = (A = 0.3362, C = 0.4093, G = 0.0743), rate matrix = (AC = 0.2419, AG = 16.2017, AT = 0.3733, CG = 0.3970, were used to create 50% majority-rule consensus trees. CT = 5.5763), c = 0.8346, proportion invariable = 0.3895. BFib5: Mod- We took a cautious approach and removed more samples el = GTR + G, base frequencies = (A = 0.3141, C = 0.1704, G = 0.2212), than indicated by visual inspection of the likelihood plots rate matrix = (AC = 1.0991, AG = 3.0395, AT = 0.5127, CG = 1.3404, to ensure that burn-in runs were not included in our con- CT = 3.6422), c = 1.0033. TGFb2: Model = GTR + G, base frequen- sensus trees (Leache and Reeder, 2002). cies = (A = 0.2419, C = 0.2152, G = 0.2068), rate matrix = (AC = 0.9510, AG = 4.6336, AT = 0.7160, CG = 1.4673, CT = 2.6666), c = 1.8795. In this study, we sequenced both mitochondrial and nuclear DNA regions with the hope that the nuclear DNA would help resolve the deeper nodes in the tree where the mitochondrial data were saturated and poten- resultant topology and node support, we ran two separate tially misinformative (see Fig. 1, Springer et al., 2001; Bayesian analyses. One analysis was run with all taxa, Moyle, 2004). We were unable to obtain the nuclear including the four taxa with missing data (complete data intron data for the four samples we extracted from set). The second Bayesian analysis excluded these taxa museum study skins (Table 1), and therefore those sam- and included only species with complete mtDNA and ples have less than half the characters of all other species nuclear intron sequences (reduced data set). A likelihood in the matrix. The effects of such incomplete matrices are analysis was run only with the reduced data set. difficult to predict, but simulations suggest the amount of To provide a statistical assessment of congruence missing data for a given taxon may be less important, in between our molecular phylogeny and the morphological terms of accurately placing it on a phylogenetic tree, than phylogeny of Burt (2004) we used the SH (Shimodaira the type and quality of the existing data (Wiens, 2003). To and Hasegawa, 1999) test. The alternative tree topology assess the effect that the missing nuclear data had on the was developed using MacClade v4.0 (Maddison and B.D. Marks et al. / Molecular Phylogenetics and Evolution 45 (2007) 23–32 27

Maddison, 2000). The SH test was used because of its con- and nuclear DNA) models. Several other studies have servative nature relative to the similar SOWH test which found support for similar partitioning schemes that has been shown to be misleading under certain conditions account for biologically meaningful differences in the (Buckley, 2002). The SH test was run with full approxima- sequence data (e.g., Brandley et al., 2005; Fyler et al., tion, 1000 bootstrap replicates, and the model of evolution 2005; Moyle et al., 2006). MrModeltest determined a gen- identified by MODELTEST. eral time reversible model framework, with c-distributed rates among sites and invariant sites (GTR + I + G) for 3. Results each codon position and the same model, but lacking invariant sites (GTR + G), for each intron. None of the 3.1. Sequence attributes three ILD tests indicated that there was significant conflict among the data partitions (ND2 vs nuclear loci: p = 0.06; The DNA sequence alignment included 33 taxa (30 ND2 vs BFib5 vs TGFb2: p = 0.58; BFib5 vs TGFb2: ingroup and three outgroup) and 2211 bp (1041 ND2, p = 0.35). Therefore, we combined ND2, BFib5, and 604 BFib5, and 566 TGFb2). Excluding the three outgroup TGFb2 data sets for all phylogenetic analyses. taxa (Momotus, Todus,andActenoides concretus), 1450 (65.5%) characters were constant, 316 (14.3%) were vari- 3.2. Phylogenetic results able but uninformative, and 447 (20.2%) were parsimony informative. Aligned ND2 sequences appear to be of mito- The genus Nyctiornis, represented in this study by Nyct- chondrial origin rather than nuclear copies. Sequences con- iornis amicta, received significant support in all analyses as tained no stop codons, overlapping fragments contained no the sister taxon of all other bee-eaters (Figs. 2 and 3). After conflicts, base composition was homogenous across taxa the initial dichotomy between Nyctiornis and the rest of the (for ND2: v2 = 39.01, df = 84, p = 0.99, TgG: v2 = 12.85, bee-eaters, Merops breweri, a sister pairing of bulocki and df = 84, p = 1.0, and Fib5: v2 = 7.19, df = 84, p = 1.0), bullockoides, and the monotypic genus Meropogon diverge codon positions contained expected relative levels of poly- in a basal polytomy at the base of the Merops radiation. morphism (3 > 1 > 2). The aligned intron sequences (BFib5 Phylogenetic analysis of the reduced data set provided a and TGFb2) contained several inferred insertions or dele- topology (Fig. 2A) with higher basal support than the com- tions (indels), but alignment of the sequences was straight- plete data set topology. The complete data set (Fig. 2B), forward for two reasons. First, indels were infrequent which includes Meropogon forsteni and Merops leschenaulti enough that they generally did not overlap, allowing without nuclear intron sequences, provides no basal sup- homologous indels to be easily identified. Second, the port after the divergence of Nyctiornis and the rest of the nucleotide sequences themselves were not highly diverged, family, resulting in a polytomy. In the reduced data set, which allowed straightforward alignment by eye and when Meropogon is excluded, the support for the positions unambiguous placement of indels. Table 2 and Fig. 1 sum- of M. bulocki/bullockoides and M. breweri dramatically marize sequence attributes and relative levels of divergence improves. for all three gene regions used in this study. Plots of pair- A clade consisting entirely of African taxa (clade A, Figs. wise divergences of the three gene regions (Fig. 1) indicate 2 and 3) was recovered in all analyses. Within this clade, the that as expected, ND2 is evolving faster than either of the two African forest specialists M. gularis and M. muelleri are nuclear introns and that among the two introns TGFb2 is highly supported as sister taxa. This pair is sister to the Afri- evolving at a faster rate than BFib5. Bayes factor analysis can savanna and forest edge species (M. hirundineus, M. provided strong evidence that a five-partition model (three oreobates, M. variegates, and M. pusillus). Relationships codon positions and each intron) was more appropriate within the savanna and forest edge clade were not fully than one partition (all data together), two partition resolved and are discussed below. (nuclear vs mtDNA), or four partition (codon positions A second well-supported clade (clade B, Figs. 2 and 3) consisted primarily of Asian/Palearctic species that share a common ancestor with a grade of three intra-African Table 2 M. albicollis M. nubicus M. malimbicus Description of mitochondrial and nuclear gene sequences migrants ( , ,and ). Within this clade are two highly supported subclades: one Gene % Characters Maximum uncorrected sequence of which unites the widespread M. orientalis sister to the divergence sedentary Southeast Asian M. viridis and M. leschenaulti. Variable Informative Within ingroup To outgroup These species are sister to a clade of mostly migratory taxa, (%) (%) including the weakly supported sister taxa, M. ornatus,an ND2 53.7 42.8 23.8a 27.7c a d intra-Australian migrant, and M. apiaster, a Palearctic TGFb2 44.5 20.3 13.2 20.8 BFIB5 34.4 13.7 10.2b 16.3c migrant, which are sister to the M. superciliosus species complex (represented here by M. superciliosus, M. persicus, a Nyctiornis amicta/Merops pusillus meridonalis. b Nyctiornis amicta/Merops malimbicus. and M. philippinus). c Momotus momota/Merops malimbicus. Parsimony analysis (MP) recovered a single most parsi- d Momotus momota/Merops philippinus. monious tree (TL = 2636, CI = 0.516, RI = 0.58; Fig. 3), 28 B.D. Marks et al. / Molecular Phylogenetics and Evolution 45 (2007) 23–32

1a. reduced data set 1b. complete data set

Merops boehmi Merops boehmi 1.0 M. muelleri mentalis 1.0 M. muelleri mentalis 1.0 M. m. muelleri 1.0 99 M. m. muelleri 1.0 M.gularis australis 100 M. g. gularis 1.0 M. gularis australis 1.0 A M. oreobates 100 M. g. gularis 1.0 A M. hirundineus Maximum likelihood bootstrap topology 100 Merops oreobates 1.0 1.0 M. pusillus pusillus M. oreobates Merops hirundineus M. p. meridonalis 80 1.0 100 M. pusillus pusillus M. pusillus pusillus M. pusillus meridonalis 94 1.0 M. variegatus lorinigi 94 M. variegatus lorinigi M. p. meridonalis M. ornatus 100 M. hirundineus M. variegatus lorinigi 1.0 M. apiaster 1.0 Merops ornatus 1.0 M. apiaster 1.0 Merops apiaster 1.0 M. persicus 1.0 M. superciliosus 100 66 1.0 Merops apiaster 0.96 M. philippinus 100 1.0 0.98 1.0 Merops persicus M. viridis viridis B 100 M. superciliosus 1.0 M. v. americanus 1.0 81 B 99 Merops philippinus 1.0 Merops leschenaulti115 1.0 Merops leschenaulti286 0.97 1.0 M. viridis viridis 1.0 1.0 1.0 1.0 100 M. v. americanus M. orientalis 76 74 91 Merops orientalis M. nubicus 1.0 1.0 M. malimbicus 97 Merops nubicus 1.0 1.0 M. albicollis Merops malimbicus 100 99 1.0 M. breweri Merops albicollis 1.0 M. bulocki 1.0 Merops breweri M. bullockoides 1.0 Merops bulocki 1.0 Meropogon315 100 1.0 Meropogon316 100 Merops bullockoides Nyctiornis amicta Nyctiornis amicta Momotus Momotus

0.4 Todus 0.42 Todus Actenoides concretus Actenoides concretus

Fig. 2. Reduced and complete data set Bayesian consensus trees. Taxa shown in bold font in the complete data set tree are removed from the reduced data set. Bayesian probabilities are above the nodes, likelihood bootstrap proportions are below the nodes on the reduced data set consensus tree. Inset box shows likelihood bootstrapping proportions for an alternate arrangement of taxa. which supports most of the same relationships seen in the 3.3. Intraspecific differentiation Bayesian and ML analyses. Topology differences exist among some of the basal nodes, but the conflicting nodes Our sampling included multiple individuals (usually dif- were not supported by bootstrapping or posterior proba- ferent subspecies) of several widespread species of bee-eat- bilities. One node received markedly different support indi- ers, allowing us a first glance at levels of intraspecific ces from each method. Under the MP and ML criterion, a variation. Within these species, uncorrected pairwise dis- clade containing M. boehmi and clades A and B received tances for the mitochondrial protein-coding ND2 gene ran- low bootstrap support. In contrast, the same node was ged from 0.04% between two individuals of M. apiaster supported by a posterior probability of 1.0. There also is to 6.5% between two individuals of M. muelleri one case where Bayesian analysis failed to provide any (M. m. mentalis from Ghana and M. m. muelleri from Equa- support for a topology supported by MP and ML boot- torial Guinea). Between these two extremes were two sub- strapping (Fig. 2A, see inset). This discrepancy involved species of M. pusillus, including M. p. pusillus from Ghana support for nodes within the African savanna and forest and M. p. meridionalis from Malawi, which differ from one edge clade including M. pusillus, variegatus, oreobates, another by 4.8% uncorrected sequence divergence. Diver- and hirundineus. Parsimony and likelihood bootsrapping gences within M. gularis and M. superciliosus are similar provided consistent support for a sister relationship to those found for M. pusillus. M. g. gularis, from Ghana, between M. hirundineus and a polytomy including M. oreo- and M. g. australis from Gabon differ by 3.7% uncorrected bates, pusillus, and variegates, whereas Bayesian analysis mtDNA sequence divergence. Finally, M. supercilious from did not provide support for any structure within this clade. Madagascar and M. philippinus from Taiwan, two taxa that We compared the Bayesian topology (Fig. 2B) with the have been alternately treated as subspecies and species, differ strict consensus tree of Burt (2004, Fig. 3b) using the SH by 3.8% uncorrected mtDNA sequence divergence. test to assess whether the plumage-based topology of Burt (2004) is significantly different from our Bayesian toplogy, 4. Discussion given the molecular data. The molecular data from this study were able to reject (p < 0.001) the topology generated Within the bee-eaters two well-supported clades can be by the morphological characters of Burt (2004). broadly defined along ecological lines: one containing B.D. Marks et al. / Molecular Phylogenetics and Evolution 45 (2007) 23–32 29

Merops boehmi and are distributed widely across the Old World. Branch- M. ornatus ing oﬀ in succession from the base of the clade are the 100 M. apiaster 11515 M. apiaster 35153 intra-African migrants M. albicollis, M. malimbicus, and, 100 M. persicus M. nubicus. Sharing a common ancestor with these 100 M. superciliosus 46361 intra-African migrants is a group of Asian and Eurasian M. philippinus 94 98 M. viridis viridis B taxa. Within this clade there are three sub-clades. First 100 M. viridis americanus is the sister pairing of the most widespread of the bee-eat- Merops leschenaulti 115 ers, M. orientalis with the Southeast Asian M. viridis and 82 100 Merops leschenaulti 286 M. orientalis M. leschenaulti. Whereas Fry (1969) suspected that M.

89 M. malimbicus orientalis was sister to the African M. boehmi, he did pro- M. nubicus pose a potential sister relationship between M. leschenaul- M. albicollis ti and M. viridis. von Boetticher (1951) recognized several 100 M. muelleri mentalis 77 M. muelleri muelleri similarities between M. orientalis, M. viridis, and M. lesch- M. gularis australis enaulti and lumped all three into a subgenus Phlothrus. 100 M. gularis gularis 88 While our data provide strong support for the arrange- M. oreobates A 88 97 M. pusillus pusillus ment of von Boetticher (1951), apart from geographic 88 M. pusillus meridonalis proximity, we see no need to split these three species from M. variegatus lorinigi 100 M. hirundineus Merops. The next well-supported sub-clade is the M. per- Merops breweri sicus/superciliosus/philippinus group; a species complex 100 Meropogon forsteni 315 that has been treated a variety of ways including one 100 Meropogon forsteni 316 widespread species with five subspecies, two species each 100 Merops bulocki Merops bullockoides with subspecific variation, and three species. Early Nyctiornis amicta authors treated it as a single widespread species with five Momotus Todus distinct subspecies distributed across Africa, Madagascar, Actenoides concretus Asia, and New Guinea and its’ surrounding islands. How- 50 changes ever, Marien (1950) noted that where M. s. persicus and Fig. 3. Single most parsimonious tree (length = 2636, CI = 0.516, RI = M. s. philippinus nest sympatrically in India there is no 0.58) shown with bootstrap values from 100 bootstrap replicates above the hybridization, and thus concluded that the two taxa are nodes. specifically distinct. Subsequent treatments (e.g., Dickin- son, 2003) follow this suggestion and further split M. s. persicus and M. s. chrysocercus from the Malagasy M. mostly sedentary species and one containing mostly migra- s. superciliosus yielding three species; M. persicus, M. tory species. Clade A (Figs. 2 and 3) is comprised entirely superciliosus, and M. philippinus. Our taxon sampling of sedentary African taxa. Its six species can be further sep- included representatives from all three of the species-level arated into two groups by habitat preference into forest taxa in this complex. M. superciliosus is sister to M. per- and savanna/forest edge species. The forest specialists M. sicus with a roughly 2% uncorrected p-distance between gularis and M. muelleri were united into their own genus, the two. That sister pairing is roughly 4% divergent from Meropiscus, based on plumage and behavioral differences its’ sister taxon M. philippinus. The level of sequence (von Boetticher, 1951). However, Fry (1969) considered divergence between the Malagasy M. superciliosus and the differences between Meropiscus and other Merops to asian M. persicus is lower than the level of divergence be misleading due to convergence between the two forest between many of the subspecific comparisons in our sam- specialists and retained them in Merops. Our data place ple (see results and discussion below). Additional behav- M. gularis and M. muelleri as sister taxa embedded well ioral, morphological, and genetic work is needed to inside the Merops radiation, providing no molecular basis clarify the status of these taxa. for splitting these two taxa from the rest of Merops. The The two species of Nyctiornis are unique among bee-eat- sister group to M. muelleri and gularis are savanna and for- ers in several ways (Fry, 1984). They differ morphologically est edge species, four African taxa whose close relation- in that they are the only bee-eaters: (1) with bi-colored ships have long been suspected as they are often treated beaks; (2) without a black eye-stripe; (3) with sexual dimor- as a superspecies (Fry, 1969): M. pusillus, hirundineus, oreo- phism (Nyctiornis amicta only); and (4) with elongated pen- bates and variegatus. These species have overlapping dant shaped throat feathers. They also differ from the rest ranges, but are rarely syntopic, occupying distinct habitats of the family vocally by producing roller-like sounds ranging from riparian grassland and woodland savanna to including a gruff ‘‘quo-qua-qua-qua-qua’’, a deep ‘‘kwow, montane forests and grasslands. All analyses recover this and hoarse guttural croaks and chuckles (Fry, 2001). All clade although relationships within the clade are not well analyses provide strong support that Nyctiornis is sister resolved. to all other Meropidae, in agreement with earlier studies In contrast to the sedentary African taxa in Clade A, documenting its uniqueness among bee-eaters (Fry, 1969, the species in clade B are all migratory to some degree 1984; von Boetticher, 1951; Burt, 2004). 30 B.D. Marks et al. / Molecular Phylogenetics and Evolution 45 (2007) 23–32

After the splitting off of Nyctiornis, our data place sev- boehmi has traditionally been linked to two other species. eral lineages in a polytomy at the base of the tree. von Boetticher (1951) grouped M. boehmi and M. albicollis Model-based analyses (Fig. 2B) recover four clades with lit- into a genus Aerops based on plumage similarities, whereas tle support for relationships among them: Meropogon Fry (1969, 1984) believed that the plumage similarities forsteni, Merops breweri, M. bullocki/M. bullockoides, and between M. boehmi and M. albicollis were convergent a large clade including M. boehmi and clades A and B. and therefore phylogenetically misleading. Fry further sug- The lack of resolution at the base of the tree seems to be gested that M. boehmi originated as a parapatric offshoot due, at least in part, to the inclusion of Meropogon in the of the widespread M. orientalis (Fry, 1969). Burt (2004) analysis. When the reduced matrix is analyzed, the basal was unable to place M. boehmi sister to any other species nodes have much higher support (Fig. 2A). This suggests with confidence. We recover two alternate phylogenetic that the relationships of Meropogon forsteni were unsettled placements for M. boehmi but neither received high sup- during Bayesian analysis and obscured the relationships of port. Model-based analyses place M. boehmi sister to clade the remaining taxa. The difficulty in resolving the position A(Fig. 3). Although this relationship is not well supported, of Meropogon may be due to the lack of nuclear data for high posterior probability unites M. boehmi with clades A that taxon. and B, separating it from the more basal lineages in the Maximum parsimony analysis places Meropogon family. In contrast, under maximum parsimony M. boehmi forsteni, the monotypic Sulawesi endemic, sister to Merops is sister to clade B, but several weakly supported nodes breweri from Africa, although this relationship is not between it and the base of the tree mean that it must be highly supported by bootstrapping. These two bee-eaters considered part of the basal polytomy (Fig. 2). Although share many morphological and behavioral similarities. the specific relationships are uncertain, this suggests that Both species are solitary inhabitants of forest edge. Mor- M. boehmi is not most closely related to either M. albicollis phologically, they share the same wing and tail colors or M. orientalis. and patterns, and both have elongated throat feathers that are often puffed out to form a beard or ruff during vocaliza- 4.1. Intraspecific variation tions (Fry, 1969). These ecological and morphological similarities have led previous authors to propose that they The intercontinental migrant M. apiaster breeds may be sister taxa (von Boetticher, 1951; Burt, 2004). throughout Europe and winters in Southern Africa. It dis- Although this relationship seems biogeographically plays very low levels of genetic differentiation between the improbable, other examples exist of sister relationships two samples in our study; one each from Kuwait and between Southeast Asian and equatorial African taxa. Malawi (0.04% uncorrected p-distance). This is in contrast For example, the River Martins Pseudochelidon sirintarae to the substantial intraspecific variation shown by the four and P. eurystomina (Zusi, 1978; Sheldon et al., 2005), and non-migratory taxa. Four out of the five species for which the Peafowls Afropavo and Pavo (Kimball et al., 1997; we had multiple individuals are non-migratory. The average Crowe et al., 2006), each have one species in equatorial uncorrected pairwise divergence within these four species African forests and other in SE Asian forests. However, was 4.6%. The greatest divergence was between two individ- despite the morphological and ecological similarities and uals of M. muelleri, from Ghana and Equatorial Guinea. the biogeographic precedent for the relationship, the They were 6.5% divergent from one another, a degree of dif- molecular data do not strongly support this sister relation- ferentiation equal to that found in taxa that are treated as ship with high bootstrap or Bayesian probabilities. More distinct species (e.g., M. ornatus to M. apiaster is 6.4%). data are needed to unequivocally assess this sister In addition to the substantial genetic break, subspecies of relationship. M. muelleri are diagnosable by the length of the central rec- Our data also have difficulty placing the sedentary Afri- trices, with M.m. muelleri having central rectrices up to can savanna species Merops bulocki and M. bullockoides, 5 mm longer than the rest of the tail and M. m. mentalis with which are well supported as sister taxa, but together are central rectrices at least 30 mm longer than the rest of the members of the basal polytomy. These taxa are allopatric, tail (Fry, 1969). Two specimens of M. gularis taken from with the range of the northern M. bulocki approaching that the same localities as the two M. muelleri were only 3.7% of M. bullockoides at the southern end of its’ distribution at divergent from one another, which may suggest that contact Lake Albert in Uganda (Fry, 1969). Because they differ in between subspecies of M. gularis occurred more recently plumage and overall size, most authors agree that these than for M. muelleri. Another possibility is that the differ- taxa are legitimate species (Irwin and Benson, 1966; Dick- ence in the level of genetic divergence within the two species inson, 2003) High divergences in both our mtDNA and is demographically driven; for example by differences in combined nuclear intron data sets (uncorrected p-distances population size. On a continental scale, the level of genetic of 7.9% and 1.9%, respectively) are consistent with recog- differentiation between populations of these African forest nizing bulocki and bullockoides as separate species. taxa is similar to that found in some other species of low- Finally M. boehmi, an inhabitant of riparian woodlands land African rainforest birds (Beresford and Cracraft, centered around Lake Malawi, Africa, has uncertain phy- 1999; Beresford, 2003; but see Roy, 1997; Bowie et al., logenetic affinities in our study. In taxonomic lists Merops 2004). The lone Asian taxon for which we have multiple B.D. Marks et al. / Molecular Phylogenetics and Evolution 45 (2007) 23–32 31 individuals is Merops viridis, an inhabitant of open country References from southeast China through the Indonesian and Philip- pine archipelagos to Java. We compared one individual of Armenta, J.K., Weckstein, J.D., Lane, D.F., 2005. Geographic variation each described subspecies. The M. v. viridis from Borneo in mitochondrial DNA sequences of an Amazonian nonpasserine: the Black-spotted Barbet complex. Condor 107 (3), 527–536. and M. v. americanus from the Philippines are 3.8% diver- Beresford, P., 2003. Molecular systematics of Alethe, Sheppardia and gent from one another, a comparable value to other subspe- some other African robins (Muscicapoidea). Ostrich 74, 58–73. cific comparisons. Beresford, P., Cracraft, J., 1999. Speciation in African forest Robins We analyzed two individuals from the monotypic inter- (Stiphrornis): species limits, phylogenetic relationships, and molecular continental migrant M. apiaster, which showed a negligible biogeography. Am. Mus. Novitates 3270, 1–22. von Boetticher, H., 1951. La syste´matique des gueˆpiers. Oiseau Rev. Fr. level of differentiation (0.04%) between birds collected in Ornithol. 5, 194–199. Kuwait and Malawi. Thus, the lone migratory taxon shows Bowie, R.C., Fjeldsa˚, J., Hackett, S.J., Crowe, T.M., 2004. Molecular low levels of genetic divergence relative to the non-migra- evolution in space and through time: mtDNA phylogeography of the tory taxa, a pattern seen in several other avian groups Olive Sunbird (Nectarinia olivacea/obscura) throughout continental (McDonald et al., 1999; Korfanta et al., 2005). Taken Africa. Mol. Phylogenet. Evol. 33, 56–74. Buckley, T.R., 2002. Model misspecification and probabalistic tests of together, these comparisons indicate that genetic diversity topology: evidence from empirical data sets. Syst. Biol. 51, 509–523. can be substantial within described species of bee-eaters, Burt, D.B., 2004. Plumage-based phylogenetic analyses of the Merops bee- especially non-migratory ones, which is consistent with a eaters. Ibis 146, 481–492. higher species level diversity than is currently recognized. Brandley, M.C., Schmitz, A., Reeder, T.W., 2005. Partitioned Bayesian Denser taxon sampling and more intraspecific studies in analysis, partition choice, and the phylogenetic relationships of scincid lizards. Syst. Biol. 54, 373–390. general are needed to gain a better understanding of genetic Crochet, P.A., Bohhomme, F., Lebreton, J.D., 1999. Molecular phylogeny structure at this level. and plumage evolution in gulls (Larini). J. Evol. Biol. 12, 1–11. Cunningham, C.W., 1997. Can three incongruence tests predict when data Acknowledgments should be combined? Mol. Biol. Evol. 14, 733–740. Crowe, T.M., Bowie, R.C.K., Bloomer, P., Mandiwana, T.G., Hedderson, T.A.J., Randi, E., Pereira, S.L., Wakeling, J., 2006. Phylogenetics, Tissue samples were kindly provided by the Field Mu- biogeography and classification of, and character evolution in, seum of Natural History, the Academy of Natural Sciences gamebirds (Aves: Galliformes): effects of character exclusion, data Philadelphia, the National Musuem of Natural History, partitioning and missing data. Cladistics 22, 495–532. UNLV Barrick Museum, D. Brent Burt, the LSU Museum Dickinson, E.C. (Ed.), 2003, third ed.. The Howard and Moore Complete Checklist of the Birds of the World Princeton University Press, of Natural Science, and the Zoological Museum at the Uni- Princeton, NJ. versity of Copenhagen. We are also grateful to the field col- Driskell, A.C., Christidis, L., 2004. Phylogeny and evolution of the lectors at these institutions for all of their effort and Australo-Papuan honeyeaters (Passeriformes, Meliphagidae). Mol. generosity. Fred Sheldon, John Bates, and an anonymous Phylogenet. Evol. 31, 943–960. reviewer provided helpful critiques of the manuscript. Farris, J.S., Kallersjo, M., Kluge, A.G., Bult, C., 1994. Testing significance of incongruence. Cladistics 10, 315–319. For collecting permits and field assistance, we thank the Farris, J.S., Kallersjo, M., Kluge, A.G., Bult, C., 1995. Constructing a Malaysian Economic Planning Unit, Sabah Museum (Jo- significance test for incongruence. Syst. Biol. 44, 570–572. seph Guntavid, Jaffit Majuakim and Albert Lo), Sabah Felsenstein, J., 1985. Confidence limits on phylogenies: An approach using Parks (Datuk Lamri Ali, Jamili Nais, and Maklarin La- the bootstrap. Evolution 39, 783–791. kim), and Sabah Wildlife Department (Datuk Mahedi An- Fry, C.H., 1969. The evolution and systematics of bee-eaters (Meropidae). IBIS 111, 557–592. dau and Augustine Tuuga), the Nature Conservation Fry, C.H., 1972. Social organization of bee-eaters (Meropidae) and Research Centre (John Mason, Larsey Ashrifie, Patrick cooperative breeding in hot-climate birds. IBIS 114, 1–14. Adjewodah, and Mavis Boateng), the Ghana Wildlife Divi- Fry, C.H., 1984. The Bee-Eaters. Poyser, Calton, Staffordshire. sion (Mike Adu-Nsiah, Joseph Amponsah, James Oppong, Fry, C.H., 2001. Family Meropidae (bee-eaters). In: Handbook of the and James Braimah), the Uganda Wildlife Authority and Birds of the WorldMousebirds to Hornbills Lynx Ediciouns, vol. 6. Barcelona, Spain. Uganda Forestry Department. R.G.M. was supported in Fyler, C.A., Reeder, T.W., Berta, A., Antonelis, G., Aguilar, A., part by the F.M. Chapman Memorial Fund. A portion Androukaki, E., 2005. Historical biogeography and phylogeny of of the research was conducted in the Ambrose Monell monachine seals (Pinnipedia: Phocidae) based on mitochondrial and Molecular Laboratory and is a contribution of the Lewis nuclear DNA data. J. Biogeogr. 32, 1267–1279. B. and Dorothy Cullman Program for Molecular System- Hackett, S.J., 1996. Molecular phylogenetics and biogeography of tanagers in the genus Ramphocelus (Aves). Mol. Phylogenet. Evol. 5, 368–382. atic Studies, a joint initiative of the New York Botanical Hegner, R.E., Emlen, S.T., Demong, N.J., 1982. Spatial-organization of Gardens and the American Museum of Natural History. the white-fronted bee-eater. Nature 298, 264–266. Data collection at LSUMNS was supported in part by Huelsenbeck, J.P., Bollback, J.P., 2001. Empirical and hierarchical the NSF (DBI-0400797 and EF-0228688). Data collection Bayesian estimation of ancestral states. Syst. Biol. 50, 351–366. at FMNH was supported by the NSF DEB-0515672 to Irwin, M.P.S., Benson, C.W., 1966. Notes on the birds of Zambia: 2. Arnoldia (Rhodesia), (37) 2, 21 pp. J.D.W. and was carried out in the Pritzker Laboratory Johnson, K.P., 1999. The evolution of bill coloration and plumage for Molecular Systematics and Evolution, operated with dimorphism supports the transference hypothesis in dabbling ducks. support of the Pritzker Foundation. Behav. Ecol. 10, 63–67. 32 B.D. Marks et al. / Molecular Phylogenetics and Evolution 45 (2007) 23–32

Johnson, K.P., Sorenson, M.D., 1998. Comparing molecular evolution in Omland, K.E., Lanyon, S.M., 2000. Reconstructing plumage evolution in two mitochondrial protein coding genes (cytochrome b and ND2) in orioles (Icterus) repeated convergence and reversal in patterns. the dabbling ducks (Tribe Anatini). Mol. Phylogenet. Evol. 10, 82–94. Evolution 54, 2119–2133. Johnson, K.P., Lanyon, S.M., 2000. Evolutionary changes in color Omland, K.E., Hofmann, C.M., 2006. Adding color to the past ancestral patches of blackbirds are associated with marsh nesting. Behav. Ecol. state reconstruction of bird coloration. In: Hill, G.E., McGraw, K.J. 11, 515–519. (Eds.), Bird Coloration, Function and Evolution, vol. 2. Harvard Kimball, R.T., Braun, E.L., Ligon, J.D., 1997. Resolution of the University Press, Cambridge, MA, pp. 417–454. phylogenetic position of the Congo peafowl, Afropavo congensis a Posada, D., Crandall, K.A., 1998. MODELTEST. Testing the model of biogeographic and evolutionary enigma. Proc. R. Soc. Lond., B 264, DNA substitution. Bioinformatics 14, 817–818. 1517–1523. Primmer, C.R., Borge, T., Lindell, J., Saetre, G.P., 2002. Single-nucleotide Korfanta, N.M., McDonald, D.B., Glenn, T.C., 2005. Burrowing owl polymorphism characterization in species with limited available (Athene cunicularia) population genetics. A comparison of north sequence information high nucleotide diversity revealed in the avian american forms and migratory habits. Auk 122, 464–478. genome. Mol. Ecol. 11, 603–612. Leache, A.D., Reeder, T.W., 2002. Molecular systematics of the Eastern Ronquist, F., Huelsenbeck, J.P., 2003. MrBayes 3. Bayesian phylogenetic Fence Lizard (Sceloporus undulatus). A comparison of parsimony, inference under mixed models. Bioinformatics 19, 1572–1574. likelihood, and Bayesian approaches. Syst. Biol. 51, 44–68. Roy, M.S., 1997. Recent diversification in African greenbuls (Pycnonot- Maddison, D.R., Maddison, W.P. 2000. MACCLADE 4: Analysis of idae. Andropadus) supports a montane speciation model. Proc. R. Soc. Phylogeny and Character Evolution, Version 4.0. Sinauer Associates, Lond., B 264, 1337–1344. Sunderland, Massachusetts. Sheldon, F.H., Whittingham, L.A., Moyle, R.G., Slikas, B., Winkler, Marien, D., 1950. Notes on some Asiatic Meropidae. J. Bombay Nat. D.W., 2005. Phylogeny of swallows (Aves. Hirundinidae) estimated Hist. Soc. 49, 151–164. from nuclear and mitochondrial DNA sequences. Mol. Phylogenet. McDonald, D.B., Potts, W.K., Fitzpatrick, J.W., Woolfenden, G.E., 1999. Evol. 35, 254–270. Contrasting genetic structures in sister species of North American Shimodaira, H., Hasegawa, M., 1999. Multiple comparisons of loglike- scrub-jays. Proc. Royal Soc. London B. 266, 1117–1125. lihoods with applications to phylogenetic inference. Mol. Biol. Evol. Moyle, R.G., 2004. Phylogenetics of barbets (Aves. Piciformes) based on 16, 1114–1116. nuclear and mitochondrial DNA sequence data. Mol. Phylogenet. Springer, M.S., DeBry, R.W., Douady, C., Amrine, H.M., Madsen, O., de Evol. 30, 187–200. Jong, W.W., Stanhope, M.J., 2001. Mitochondrial versus nuclear gene Moyle, R.G., Cracraft, J., Lakim, M., Nais, J., Sheldon, F.H., 2006. sequences in deep-level mammalian phylogeny reconstruction. Mol. Reconsideration of the phylogenetic relationships of the enigmatic Biol. Evol. 18, 132–143. Bornean Bristlehead (Pityriasis gymnocephala). Mol. Phylogenet. Evol. Swofford, D.L., 2002. PAUP*. Phylogeneitc Analysis Using Parsimony (* 39, 893–898. and other methods), Version 4.0b10. Sinauer Associates, Sunderland, Mundy, N.I., Unitt, P., Woodruff, D.S., 1997. Skin from feet of museum MA. specimens as a non-destructive source of DNA for avian genotyping. Weckstein, J.D., 2005. Molecular phylogenetics of the Ramphastos Auk 114, 126–129. toucans. Implications for the evolution of morphology, vocalizations, Nylander, J.A.A., 2004. MrModeltest v2. Program distributed by the and coloration. AUK 122, 1191–1209. author. Evolutionary Biology Centre, Uppsala University. Wiens, J.J., 2003. Missing data, incomplete taxa, and phylogenetic Nylander, J.A.A., Ronquist, F., Huelsenbeck, J.P., Nieves-Aldrey, J.L., accuracy. Syst. Biol. 52, 528–538. 2004. Bayesian phylogenetic analysis of combined data. Syst. Biol. 53, Zusi, R.L., 1978. Remarks on the generic allocation of Pseudochelidon 47–67. sirintarae. Bull. Brit. Orn. Cl. 98 (1), 13–15.